WHAT NEW IN X10.29 THU

X 上的 AI 合集档案 · 22 个合集 · 60,636 条

今天新增personal-agentIMAGE

滑雪住宿 MCP 服务器上线,让个人助理直接查实时房源

MV Travel 发布 MCP 服务器,Claude、Cursor、Gemini 等助手可查询欧洲滑雪度假村的实时可订房源,作者还展望了助理与商家助理互相议价的未来。

Dr. Tali Režun
@talirezun6 小时前收录原推发布于 2026.10.28
滑雪住宿 MCP 服务器上线,让个人助理直接查实时房源
Dr. Tali Režun @talirezun原推发布于 2026.10.28
在 X 上打开原推

Your next ski holiday might be planned by two AI agents talking to each other. Today I shipped a new version of the @mountvacation, now MV Travel MCP server. It connects AI assistants like Claude, ChatGPT-style coding agents, Cursor and Gemini directly to MV Travel’s live accommodation data for ski and holiday resorts across Europe. What’s new: 🏔️ Smarter place search. Katschberg, Ischgl, Flachau and dozens more now return real, bookable offers instead of “not found”. 🧭 Cross-border safety. Ask for Sestriere and you get Italy, not a French hotel from the same ski area. ⚡ A one-line installer. It finds the AI apps on your computer (14 so far, including Claude, Cursor, Codex, Gemini and VS Code) and connects them all in about a minute. Why this matters more than a feature list MCP (Model Context Protocol) is quickly becoming the standard way an AI model plugs into real systems. For tourism, that changes the interface. Travellers no longer search a website. They ask in plain language: “7 nights in March, two adults and a 9-year-old, ski-in/ski-out, under €2,500”. The agent then queries live inventory, compares prices and returns booking links. That’s useful today. The bigger shift is what comes next: the agent-to-agent economy. Picture this. Your personal agent knows your dates, budget, family and the slopes you like. It talks to MV Travel’s agent, which is connected through MCP to their API with real-time availability and prices. The two negotiate, compare options and come back to you with a shortlist. Nothing gets booked until you approve it. For that to work across companies, agents need more than an API. They need: 🔐 a private channel that the platforms in between can’t read 🤝 a structured way to propose and accept terms ✅ the human owner approving anything binding 📜 optional proof of exactly what was agreed, and when That is why I’m also building AgentLink: encrypted, provable exchange between AI agents, even when they run in different tools and belong to different people. It’s open source and pre-alpha, and runs on test networks today. MV Travel MCP gives an agent access to the data. AgentLink gives agents a trusted way to talk and agree. Together they’re a small, working preview of how travel, and much more, will be bought and sold. 🔗 MV Travel MCP: http://github.com/talirezun/MV-MCP-server 🔗 AgentLink: http://github.com/4thtech/agentlink #AI #MCP #AgenticAI #Tourism #TravelTech #AIAgents #AgentLink

x.com/talirezun/status/2115395980669436341采集于 2026.10.29 10:44